J. Non-Newtonian Fluid Mech.

166 (2011) 3241

Contents lists available at ScienceDirect

Journal of Non-Newtonian Fluid Mechanics

journal homepage: www.elsevier.com/locate/jnnfm

Bubble cluster formation in shear-thinning inelastic bubbly columns

J. Rodrigo Vlez-Cordero, Roberto Zenit
Instituto de Investigaciones en Materiales, Universidad Nacional Autnoma de Mxico, Circuito Exterior s/n, Apdo. Postal 70-360 Mxico D.F. 04510, Mexico

a r t i c l e

i n f o

Article history:
Received 13 July 2010
Received in revised form
27 September 2010
Accepted 8 October 2010

Keywords:
Bubbly ows
Non-Newtonian
Shear-thinning
Clusters

a b s t r a c t
The mean rise velocity of bubble swarms ascending in shear-thinning uids was experimentally measured
in a rectangular bubble column. Great care was taken to produce nearly mono-dispersed bubble swarms
and to use shear-thinning uids with negligible elastic effects. In this manner, it was possible to isolate the
effect of the hydrodynamic interaction between bubbles in the column caused by the thinning behavior
of the liquid. It was found that the mean rise velocity of the bubbles was larger than that of an individual
bubble, in accordance with previous studies. The magnitude of the swarm velocity was found to be greatly
inuenced by the appearance of bubble clusters. The bubble clusters, which appeared for certain values of
the ow index and bubble diameter, were found to have a very different structure from those observed
in Newtonian liquids. Furthermore, it was found that the appearance of clusters produced a dramatic
increase of the agitation within the column. A set of conditions was identied for the appearance of
bubble clusters in shear-thinning inelastic bubbly columns.
2010 Elsevier B.V. All rights reserved.

1. Introduction
One of the greatest challenges in uid dynamics is to understand and predict the behavior of multiphase ow systems.
This is a task that is widely justied due to the occurrence of
gasliquid/gasliquidsolid contactors in almost every branch of
the chemical and metallurgical industries. In particular, bubble
columns are used in many chemical processes which involve oxidation, chlorination, polymerization and hydrogenation [13], to
name a few. Bubble columns are also used as a central unit operation for primary and secondary metabolites production and for
several downstream processes for product recovery such as the
separation by adsorption in bubble ows [1,4,5]. In spite of the
mechanical simplicity that characterizes bubble columns, which
grant them with operative and cost benets [6], their basic engineering design is not a simple task. Given a certain chemical process
and liquid phase properties, an engineer has to implement the optimum column geometry, sparger conguration and power input
(supercial gas velocity Ug ) to satisfy the uptake rate of the ongoing
chemical reaction, in addition to achieving the optimum mixing and
heat transfer properties between phases. Such estimations require
the knowledge of functional relations among the central operational variables such as the mean bubble diameter d b , mean bubble
SW and gas fraction g . Some authors have also studied
velocity U
the development of the liquid velocity and its variance as a measure of the amount of agitation or pseudo-turbulence conferred by

Corresponding author. Tel.: +52 55 5622 4593; fax: +52 55 5622 4602.
E-mail address: zenit@servidor.unam.mx (R. Zenit).
0377-0257/$ see front matter 2010 Elsevier B.V. All rights reserved.
doi:10.1016/j.jnnfm.2010.10.003

the dispersed phase [7,8]. Deeper understanding of bubblebubble

and bubbleliquid interactions require the use of computational
uid dynamics in which one can easily choose and manipulate the
eld variables [911].
Although many industrial liquids which comprise solutions of
low molecular weight can be considered Newtonian-like uids, an
increasing number of solutions with high molecular weights and
internal structure are being used that have non-Newtonian behavior such as variable viscosity and memory effects. Polymer solutions
and melts, liquid crystals, gels, suspensions, emulsions, micellar
solutions, slurries and foams enter into this non-Newtonian category [10,12]. The study of non-Newtonian effects in the behavior
of bubble ows is, therefore, of fundamental importance.
Experimental and numerical studies in two phase ows (including sedimentation and bubble ows) have shown that a group of
bodies moving through a non-Newtonian uid tend to form aggregates or clusters [11,1315]. Such aggregates are more dense in
terms of the number of bubbles embedded in the bubble cloud
than the clusters that have been described in Newtonian ows at
high Reynolds numbers [16,17]; actually, we will show in this study
that bubbles rising in shear-thinning uids group with each other
forming aggregates of a certain size. Bubble clustering cause bubble coalescence and a premature transition to the heterogeneous
or churn-turbulent ow in non-Newtonian uids [18,19]. Although
there has been a number of studies of particle clustering and orientation of lengthened bodies in sedimentation [15,14,22], little has
been reported for the case of bubbly ows. The aim of the present
work is to study the formation of clusters of mono-disperse bubbles in power-law shear-thinning uids, leaving aside, as much
as possible, the elastic effects. We support the discussion of our

J. Rodrigo Vlez-Cordero, R. Zenit / J. Non-Newtonian Fluid Mech. 166 (2011) 3241

experimental observations with the results of a companion paper

[23] which analyzes the hydrodynamic interaction between bubble
pairs ascending in shear-thinning uids, both experimentally and
numerically.

33

upper capillary

10.0cm

5.0cm

2. Background
Most of the published papers on non-Newtonian bubble
columns have focused their attention on the development of
the gas fraction, g , in terms of the supercial gas velocity, Ug .
Many authors have used CMC (carboxymethylcellulose) solutions
(ow index n > 0.5) [18,19,2427] in coalescence conditions and
worked with relatively large values of the supercial gas velocity (Ug > 2 cm/s) [2527] such that the churn-turbulent ow was
generally achieved.
Buchholz et al. [18] reported that the mean bubble rise veloc SW measured in a single stage column was higher than the
ity U
single bubble velocity USI . Additionally, this difference increased
with the thinning behavior of the uid. However, the changes of
SW /USI ratio with the gas fraction were not reported explicitly.
the U
Schumpe and Deckwer [19] found that the homogeneous bubble
regime in CMC solutions could only be achieved having Ug values
below a critical one (0.5 cm/s), this value being a function of the
effective viscosity. Haque et al. [26,27] reported an estimation of
the mean bubble diameter and velocity of bubble swarms rising in
viscoelastic uids. One of the key assumptions in this estimation is
that the mean bubble velocity values are equal to the single bubble velocity, which is not always true. As far as we know, there are
no previous studies of mono-dispersed bubble swarms in shearthinning uids at low supercial gas velocities (Ug < 1 cm/s) where
the bubbly ow regime can be observed. Having a monodispersed
bubble distribution allows us to isolate the effects caused solely by
the hydrodynamic interaction among bubbles.
Many relevant contributions to the subject have arisen from
numerical and analytical studies. Several methods such as
variational principles [28], perturbation methods [29] and approximations to the Newtonian behavior [30] have been used to solve
numerically the rising of a bubble or bubble swarms through
shear-thinning uids. Bubble interactions were simplied using the
Happel cell model [31], hence no direct bubble interactions have
been studied previously. Only Radl et al. [11] have fully resolved
the ow eld around a group of bubbles. Bhavaraju et al. [29]
reported that for creeping ow conditions the drag coefcient Cd
decreased with the increase of the thinning behavior for bubble
swarms, which is the opposite of what happens for single bubbles.
Ascending isolated bubbles have a larger Cd as the thinning condition increases [27,30,32,33]. In more recent studies, which took
into account the inertial forces in the momentum equation [10,34],
this increase in the drag force for the single bubble cases occurred
only below a critical Reynolds number (approximately from 3 to
10). Gummalam and Chhabra [28,35] reported that for the creep SW /USI ratio
ing and high Re numbers (100 < Re < 300) ows, the U
increases for gas fractions below 0.3 and ow index values below
0.5. Above this gas fraction value the velocity ratio decreases. Therefore, a maximum in the velocity ratio was observed for g 0.3.
These authors explained that such behavior is due to the constrain of the ow lines (as the effective volume for each bubble
is decreased as the gas fraction increases) resulting in an increase
of the shear rate which in turn results in lower apparent viscosities.
A positive slope of the velocities ratio curve means that the viscosity gradients effects surpass the hydrodynamic hindrance effects.
SW /USI curve is commonly negative in NewtoThe slope of the U
nian bubbly liquids with large Reynolds and small Weber numbers
[17,36]. It is important to point out that the theoretical results of
Refs. [28,29,35] have not been compared with experiments yet.

lower capillary
(gas chamber side)
Fig. 1. Scheme of a capillary bank.

Radl et al. [11] directly simulated the interactions of bubbles (up to

9) rising in shear-thinning conditions. They showed that the bubble
interactions are signicantly enhanced when viscosity gradients
are present. They also identied mini-bubble clusters that travel
faster than the single bubble cases. This feature will be discussed
in depth in the present investigation.
SW /USI values obtained in the
In this work, we compared the U
bubble column at different gas fractions and thinning conditions
with the theoretical ones [28,35]. We also conducted an analysis
of the size of bubble clusters and of the bubble velocity variance.
Finally, we identied the hydrodynamic conditions for which bubble clusters are formed.
3. Experimental setup
3.1. Bubble column and capillary banks
A rectangular channel with 5 10 160 cm3 equipped with a
gas chamber (7.5% of the channel height) was used, similar to
that used by Martnez-Mercado et al. [8]. Pure nitrogen was introduced to the bubble column using a needle valve (Cole-Palmer
1682 ml/min MAX, stainless steel ball). The supercial gas velocity Ug
range was between 0.09 and 0.6 cm/s. The gas volume fraction, g ,
was obtained using the relation g = H/H, where H is the liquid
height without gas (140 cm) and H the height difference produced
by the introduction of the gas in the column.
Three capillary banks were constructed in order to produce different bubble sizes (db ). The internal diameter of the capillaries
(Dcap ) was selected considering the equilibrium condition between
buoyancy and surface tension forces:
3

Dcap =

db g
6

(1)

where  is the liquid density, g the gravity and  the surface tension.
In order to avoid the generation of gas jets with variable volume,
the hydraulic resistance through the capillaries should be large
such that the bubble volume depends mainly on the buoyancy and
surface forces and lesser in the gas ow rate [37]. The equivalent
capillary lengths necessary to achieve such hydraulic resistance are
sometimes of the order of meters; instead of installing long capillaries, a second capillary with a smaller inner diameter was inserted
to the main capillary but the bubble formed on the larger diameter
end. This arrangement provides the sufcient pressure drop to produce individual bubbles. As we can see in Fig. 1, identical capillaries
were placed in an acrylic perforated plate ordered in an hexagonal arrange. The design parameters of the three capillary banks are
summarized in Table 1.
3.2. Fluids
The uids used in this study fullled two conditions: (1) show a
shear-thinning power-law behavior with negligible elastic proper-

34

J. Rodrigo Vlez-Cordero, R. Zenit / J. Non-Newtonian Fluid Mech. 166 (2011) 3241

Table 1
Design parameters of the capillary banks. Dcap , internal diameter of the capillaries; N, number of capillaries; d b , experimental mean bubble diameter; , spacing between
capillaries; wall , spacing between capillaries and the wall; L, capillary length.

1
2
3

Dcap (mm)

d b (mm)

/d b

wall /d b

L (cm)

0.2
0.6
1.6

63
27
16

2.15
3.15
4.20

2.8
2.7
2.7

6.7
4.9
3.7

4.3
8.0
8.0

ties and (2) allow the formation of mono-dispersed bubbly ows,

that is, with a narrow dispersion of d b .
It was found that xanthan gum solutions in a water/glycerin
mixture fullled the above conditions. These solutions are more
stable in the presence of MgSO4 (which was added to delay coalescence [38]) than carbopol, CMC or guar gum solutions. The
preparation of the solutions was as follows: rst the xanthan gum
(Keltrol E 415, KELCO, USA) was dissolved in water (at 55 C), then
the salt (0.04 M of MgSO4 , 2500-01 J.T. Baker) was added and nally
the glycerin. The solutions were left in repose for 24 h before experiments. The rheological measurements were done in a rheometer
(TA Instruments AR1000N) with a cone-plate geometry (60 mm, 2 , a
gap of 65 m). The surface tension measurements were performed
with a DuNouy ring (diameter of 19.4 mm, KSV Sigma 70). All the
solutions were stirred before the surface tension measurement. The
temperature of the room was 23 C. The physical properties of the
solutions are summarized in Table 2; the ow curves are shown
in Fig. 2. The data contained in this gure includes measurements
performed one and two weeks after the preparation of the uids.
At a shear rate of 10 s1 all the uids have a viscosity of around
one hundred times that of water. The shear-thinning uids have
a power-law behavior in almost all the shear rate interval; a mild
plateau can be observed for shear rates below 0.1 s1 . This indicates
that the polymer chains are gently stretched by a ow eld, having a short range of linear viscoelasticity. Within this short linear
range, the elastic modulus G becomes bigger than the loss modulus
G only at high frequencies (>100 s1 ) as can be seen in Fig. 3. Both
G and G were obtained by a Fourier Transform (FT) of the stress
relaxation curves measured with an ARG2 rheometer of controlled
stress [39]. The normal forces (rst normal stress difference) of
the thinning uids, measured with the cone-plate geometry, reach
a maximum value of 7 Pa at 100 s1 but decreased as the shear
rate increased (data not shown). This behavior contrasts with the
one shown by a viscoelastic uid (a 0.2% polyacrylamide solution
can reach a normal stress value of 100 Pa for the same range of

For the measurements of the mean bubble diameters, d b , and

SW , a high speed camera (MotionScope
the mean bubble velocities, U
PCI 8000s) was used. A recording rate of 500 frames/s and a shutter
speed of 1/1000 was used. The camera was positioned 20 cm below
the liquid surface facing the larger side of the bubble column. Two
different videos were obtained per experiment: one for the d b measurements, where the camera was placed 120 cm from the bubble
column (using a 105 mm objective lens) and another for the bubble velocity measurements, where the camera was placed a little
bit farther, 180 cm from the bubble column (using a 60 mm objective lens). Diffuse back light was used, with a light source placed
at 90 from the camera direction and reected by a panel. The
video frames were converted to binary format and analyzed using a
conventional image processor software (Matlab ). The equivalent
mean bubble diameter d b was calculated using the measurement
of the short and long diameters of the elliptic bubble projections:
1/3
2
d b = (d MAX
d MIN )

10

10

10
10

(2)

where d MAX is the larger bubble diameter and d MIN the shorter
bubble diameter. In some bubbly ows an equivalent mean cluster
diameter was also measured. Although clusters shapes are highly
irregular, an estimation of its size was made using Eq. (2), now tak-

(mPas)

10

3.3. Bubble diameters and velocities measurements

G,G(Pa)

10

shear rates) or an upper-convected Maxwell uid where the normal stress grows as a function of  2 ; clearly the behavior of our
uids is very different. With these uids (and with their rheological characterization) we were able to isolate, as much as possible,
the thinning effects from the elastic ones over a wide range of characteristic ow times. We can expect then that the hydrodynamic
behavior of bubblebubble interactions is mainly attributed to the
shear-thinning behavior.

10

2
1

10

10

10

10

frequency (1/s)

10 1
10

10

10

10

10

10

shear rate (1/s)

Fig. 2. Flow curves of the test uids. : apparent viscosity, ( ) Newtonian uid,
() n = 0.85, () n = 0.76, () n = 0.55, () n = 0.32.

Fig. 3. Dynamic moduli of three shear thinning solutions. Filled symbols: elastic
modulus G ; empty symbols: loss modulus G ; () n = 0.85, () n = 0.55, () n = 0.32,
() a polyacrylamide reference solution (0.04% in 80% glycerin/water with 0.04 M
MgSO4 ). For the estimation of G and G the procedure followed by Calderas et al.
[39] was used.

J. Rodrigo Vlez-Cordero, R. Zenit / J. Non-Newtonian Fluid Mech. 166 (2011) 3241

35

Table 2
Physical properties of the uids: , density; , surface tension; , viscosity; k, consistency index; n, ow index. The percentages of liquid mixtures are given in volume terms,
the percentages of the xanthan gum solutions in weight terms. In all the uids 0.04 M of MgSO4 was added to delay bubble coalescence.
Fluids

 (kg/m3 )

 (mN/m)

 or k (mPa sn )

Newtonian: 83% glycerin/water

0.02% Xanthan gum in 75% glycerin/water
0.035% Xanthan gum in 70% glycerin/water
0.1% Xanthan gum in 60% glycerin/water
0.23% Xanthan gum in water

1214.6
1193.1
1179.5
1152.1
997.5

61.9
63.0
62.0
65.0
66.5

104.2
118.7
143.4
360.0
574.9

1.0
0.85
0.76
0.55
0.32

ing the longer and shorter diameters of the cluster. The reported
SW values were the average of the vertical component of the
U
velocity vector. At least 100 measurements were done for each
experimental condition. When bubble aggregates appeared, we
measured the velocity of individual bubbles located at the periphery of the clusters; these were more easy to recognize than the
bubbles located in the interior of the cluster. The uids were discarded after one week of experimental work. The whole column
was disassembled and cleaned at the end of the day.
The velocity and size of isolated bubbles were also measured
covering a bubble volume range from 0.1 to 100 mm3 . Such
measurements were done in a cylindrical column with an inner
diameter of 9 cm equipped with a bubble dispenser similar to the
one used by Soto et al. [40].

4. Experimental results
4.1. Single bubbles
The values of the drag coefcient, Cd , of the individual bubbles
as a function of the Reynolds number, Re, are shown in Fig. 4. The
Reynolds number and drag coefcient were dened as:
USI db


Re =

(3)

where
=k

 2U n1

10

SI

(4)

db

and
Cd =

4 db g
3 U2
SI

(5)

In addition to the common functional dependence of the drag coefcient on the Reynolds number in the laminar ow regime, we can
also observe that the Cd values found for the thinning uids are
higher than those found for the Newtonian uid. This increase of
the drag coefcient with the thinning behavior has been already
reported by theoretical [30,29], experimental [29,27] and numerical [23] studies for the creeping and small but nite Reynolds
regimes. Note that for the Newtonian uid the Oseen wake is
already seen at Re 7. With these measurements, the accuracy of
our experimental results was veried. Although we cannot state
that our uids are totally free from elastic effects, i.e. the ow
around a bubble is complex so memory effects can rise from extensional ow; the fact that we did not observe any signal of a velocity
jump discontinuity supports the statement that the elastic effects
were very small indeed.
4.2. Bubbly ow regime and cluster formation
As pointed out by other authors [18,19] the transition from bubbly to the heterogeneous regime, which is characterized by the
presence of multi-dispersed bubble swarms, can be identied by
a change in the slope of the gas volume fraction as a function of the
supercial gas velocity. The slope of the curve changes (decreases)
as a consequence of the change of the bubble volume. In Fig. 5
the gas fraction value is shown as a function of the supercial gas
velocity for the Newtonian and a thinning uid (n = 0.85) for several
bubble diameters. For both cases the mean bubble size produced
by each capillary bank was practically the same.

4
3.5

2.1mm

3
10

3.1mm

Cd

g (%)

2.5

10

2
1.5
4.2mm

1
0.5
0

10 1
10

10

10

10

Reynolds
Fig. 4. Drag coefcient as a function of the Reynolds number for the single bubbles.
(+) Newtonian uid, () n = 0.85, () n = 0.76, () n = 0.55, () n = 0.32, () Stokes
prediction, ( ) Hadamard prediction, (. . .) Oseen correction for a uid sphere
Cd = 16/Re + 2.

0
0.05

0.1

0.15

0.2

0.25

0.3

superficial gas velocity Ug (cm/s)

Fig. 5. Gas fraction values as a function of the supercial gas velocity. Filled symbols:
Newtonian uid, empty symbols: n = 0.85 uid, () d b = 2.1 mm, () d b = 3.1 mm, ()
d b = 4.2 mm.

36

J. Rodrigo Vlez-Cordero, R. Zenit / J. Non-Newtonian Fluid Mech. 166 (2011) 3241

Fig. 6. Bubbly ow pictures taken at different gas fraction values for the Newtonian uid and the n = 0.55 thinning uid. The magnication of the images is approximately the
same. The image size is about 5 3 cm2 . The bubble size (2.1 mm) correspond to the one produced by the smallest capillary diameter. The Reynolds number was computed
using the mean bubble diameter and velocity of the bubbly ow. The Etvs (Eo) number is dened below.

of deformation, such periodic movement seen in Newtonian uids

disappear as soon as inertia (Re > 0.2) or the number of particles
(around 7) are increased. Fig. 6 clearly shows that clustering in
thinning uids appears above these limiting factors.
4.3. Bubble size distribution
Fig. 7 shows the histogram of the bubble diameter for the Newtonian uid and the n = 0.76 uid; the plot shows data taken for each
35
2.1mm

3.1mm

4.2mm

30

25

frequency

For the Newtonian case (lled symbols) the bubbles with the
smaller d b produce higher gas fraction values, as expected. The
transition to the heterogeneous ow can be seen by the change of
the linear slope of the curves for the three different bubble diameters. The g values obtained for the thinning uid are similar to
the Newtonian case for 3.1 and 4.2 mm bubble diameters; for the
smaller bubbles (2.1 mm) the gas fraction values are clearly below
the Newtonian ones; this is due to the presence of bubble clusters,
as we will show below. Note that with the other bubble diameters,
where no bubble cluster were found, the transition to the heterogeneous regime was observed at higher values of Ug with respect
to the Newtonian uid. All the measurements reported hereafter in
this paper were taken before the transition to the heterogeneous
regime.
Fig. 6 shows images of the bubbly ows produced with the
present setup. The images correspond to the Newtonian uid and to
the shear-thinning uid with n = 0.55 for the smallest bubble size.
The formation of large bubble aggregates in the thinning uid is
largely evident (Fig. 6c and d). The difference with the Newtonian
case (Fig. 6a and b) is striking.
While the orientation of such clusters is mainly horizontal,
spheroidal clusters and small bubble chaining can also be seen. It is
worthwhile to mention that the clusters are not static while rising
through the uid; on the contrary, they have a dynamic structure. A
careful observation reveals that the bubbles move in toroidal trajectories, rising in the center and descending on the exterior part of the
cluster. As the gas fraction increases, just before the regime transition, the bubbles embedded in such clusters inevitably coalesce
and form large cap bubbles.
A similar toroidal or periodic movement has also been observed
in the case of settling particles forming clusters at low Re
[20,21]. Nevertheless, besides the difference between the degree

20

15

10

0
1

1.5

2.5

3.5

4.5

bubble diameter (mm)

Fig. 7. Histograms of the equivalent mean bubble diameters d b for the Newtonian
() and n = 0.76 ( ) uids produced by the three capillary banks; 15 classes were
used to sort the bubble diameters. The frequency refers to the number of values
found in each class.

J. Rodrigo Vlez-Cordero, R. Zenit / J. Non-Newtonian Fluid Mech. 166 (2011) 3241

37

capillary bank. In both cases the curves show a narrow distribution

of the bubbles diameters produced by the capillary banks indicating that the formation of nearly mono-dispersed bubbly ows was
accomplished. In addition, it can be seen that the bubble diameter
distributions are similar for both cases, indicating that the different uids used in this work allowed the formation of bubbles with
similar mean bubble diameters. The same behavior was observed in
the other thinning uids; the only exception was the n = 0.32 uid
with the smallest capillary diameter for which the mean bubble
diameter was 36% higher than all the other cases.
4.4. Mean bubble velocity and drag coefcient
Fig. 8 shows the ratio of the mean bubble velocity with the single
bubble velocity as a function of the gas fraction; the three bubble
sizes produced in the bubble column are included. The theoretical
values reported by Gummalam and Chhabra [28,35] for creeping
and high Re ows (100 < Re < 300) are also shown for their lowest
ow index values and for their Newtonian results.
The experimental and theoretical results agree in that the
SW /USI curve has a positive slope for small gas fraction, in contrast
U
to the result in Newtonian uids. This means that, unlike the Newtonian ows, the hydrodynamic hindrance produced by bubbles
interactions does not produce a reduction of the bubble velocity
with g ; in the shear-thinning cases this effect is compensated by
the reduction of the apparent viscosity produced by the local shear
rate values. Hence, the Happel cell model used by Gummalam and
Chhabra [28,35] captures the basic nature of these ow types. This
fact also agrees with the early results of Buchholz et al. [18].
The values of the gas volume fraction for which the maximum
SW /USI value was observed in the experiments were, neverof the U
theless, much smaller than the theoretical ones. In fact for g > 1%
the heterogeneous regime is observed, that is why measurements
at higher volumetric ow rates or gas fractions (where the maximum theoretical values were found) are practically impossible to
attain in a bubbly ow regime.
Additionally, the experimental values of the velocities ratios
were found to be much larger than the theoretical ones. As the
bubble size decreases, the difference between the theoretical and
experimental values increases (up to nine times for the case of the
n = 0.57 uid, see Fig. 8c). Hence, the Happel cell model is not sufcient to explain the whole phenomenon. The effect which is not
accounted for in this model is the appearance of bubble clusters
(like those shown in Fig. 6c and d). These clusters are responsible
SW /USI values found in the experiments. The shape of
for the high U
the curves in Fig. 8 indicates that the clusters grow up to a certain
size just before the transition to the heterogeneous regime.
The last important difference is that while in the theoretical
results the velocity ratio increases monotonically with the thinning behavior (decreasing the ow index value), the experimental
results do not necessarily follow this trend. For instance, in Fig. 8a
and b the curves of the uids with n = 0.85 and 0.32 are closed to
the theoretical ones while the uids with n = 0.55 and 0.76 are far
from them due to the formation of clusters.
The values of the drag coefcient (in terms of a drag correction factor X = Cd Re/24) were also reported in the theoretical work
of Gummalam and Chhabra [28]. Their results showed that the Cd
increases with the gas volume fraction for n > 0.4, as is commonly
observed, but decrease for index ow values below 0.4. In [35] the
Cd values for Re = 50 were also reported. These authors mentioned
that the Cd increases with the gas volume fraction, obtaining again
the common trend found in [28]. In a recent numerical work of
the same group [10], where the Happel cell model was also used,
the Cd values were found to increase with the gas fraction at the
same n value although such increase was found to be negligible for
g < 0.001. In the present work an estimation of the Cd values of the

SW /USI ratio as a function of the gas hold up g for the three mean bubFig. 8. U
ble diameters. () n = 0.85, () n = 0.76, () n = 0.55, () n = 0.32, theoretical values
[28,35]: () creeping ow n = 1.0, ( ) creeping ow n = 0.3, (--) 100 < Re < 300
n = 1.0, ( ) 100 < Re < 300 n = 0.2. The dotted horizontal line at 1.5 divides the free
bubble and cluster formation regimes.

bubbly ows were also obtained for low gas volume fractions. We
computed the Cd values using Eq. (5) with the values of the mean
bubble diameter and velocity of the ow. For comparison, we also
calculated the Cd using the constitutive equation proposed by Ishii

38

J. Rodrigo Vlez-Cordero, R. Zenit / J. Non-Newtonian Fluid Mech. 166 (2011) 3241

Table 3
Drag coefcient values for different thinning uids and gas volume fractions. In
order to calculate Cd (ReSI ) of Eq. (6) the respective bubble diameter obtained in the
bubble column was used. The data of Kishore et al. [10] corresponds to the Cd values
found for n = 0.6 and 0.8 and a gas fraction of 0.001.
n

Re

Cd

Cd Eq. (6)

0.32

0.001
0.002
0.003
0.001
0.003
0.001
0.002
0.007

10.3
10.9
10.9
5.2
5.8
3.6
4.2
4.3

3.9
3.6
2.8
2.6
2.3
4.2
3.3
2.8

4.2
3.9
3.6
6.0
5.3
8.0
6.5
6.0

0.55
0.76

Cd Ref. [10]

4.0
5.5

and Zuber [41]:

 U 2 
SI

Cd = Cd (ReSI )

USW

1 g

(6)

where the functional dependance of Cd (ReSI ) was obtained using the

single bubble data for each uid (Fig. 4). This equation was proposed
considering a multiparticle system moving in one dimension in an
innite medium (absence of walls). The results are shown in Table 3.
It can be observed that the drag coefcient actually decreases as
the gas volume fraction increases for the same uid and Re number,
unlike the general trend found by Kishore et al. [10]. Such decrease
of the Cd is directly related to cluster formation and growth. Unlike
the Newtonian uids, where the Cd increases with the gas fraction,
in bubbly shear-thinning uids it decreases. It can be seen that Eq.
(6) and the numerical results of Kishore et al. [10] both overestimate the values of Cd because they do not consider interactions or
associations between bubbles.
4.5. Cluster size
The equivalent cluster diameter d c for one of the thinning uids (n = 0.76) is presented in Fig. 9 as a function of the gas volume
fraction. In this plot the d c values produced by the smallest and
largest capillaries are shown both in dimensional and dimension3
less form (scaled by the bubble diameter d b ). The ratio (d c /d b ) was
used because it is directly proportional to the number of bubbles
in a cluster. In this particular uid clusters were observed for the
three bubble sizes tested, that is, bubble clusters are formed no
matter what the value of the d b is. Also in this uid we had the

Fig. 9. Mean equivalent cluster diameter d c as a function of g for the n = 0.76 uid.
In (b) the values are divided by the mean equivalent bubble diameter d b and elevated
to the third power. () d b = 2.1 mm, () d b = 4.2 mm. The lines are only visual aids.

4.6. Bubble velocity variance

thinning uid with n = 0.85 were determined for the three bubble
diameters; these results are shown in Fig. 10. For this uid no cluster
formation was observed for the medium and large bubbles; clusters appeared only with the smaller bubble size (see Fig. 8). As can
be observed in Fig. 10, the normalized bubble velocity variances for
the medium and large bubbles (4 < Re < 14) are in agreement with
the values obtained by Martnez-Mercado et al. [8] for a Newtonian
uid with similar Re values (10 < Re < 30). In the case of the smaller
bubbles (2.1 mm, with 1.7 < Re < 3.4), where clusters are formed,
2 values surpass by nearly one order of magnitude the
the Tb /U SW
values achieved by the unclustered cases. This cluster-uctuation
relation seen at low Re numbers is very different from that expected
in potential ows [36]. A common feature in high Re ows is that
clustering is suppressed by bubble velocity uctuations [16]. In
the present study, uctuations are enhanced by the clusters. Such
difference adds to the already large list of the contrasting behavior between inertial and viscous dominated ows. Similar Tb /U 2

We can expect that bubble cluster formation will affect the agitation levels in the liquid; hence, a larger bubble velocity variance
will be present in comparison with ows where no clusters are
found. To verify this argument the bubble velocity variance was
2 ) for the
measured. The normalized values of the variance (Tb /U SW

values and behavior were obtained for the other thinning uids.
It is interesting to note that the values of the bubble velocity variance found here for a range of 1.7 < Re < 3.4 and thinning conditions
are of the same magnitude as the liquid velocity variance found by
Cartellier and Rivire [42] for similar Re values (0.66 < Re < 1.48) and
gas fractions in Newtonian uids, although in their experiments the

opportunity to take more data of the cluster diameters before the

heterogeneous regime occurred. For the n = 0.55 uid the transition to this regime occurs at lower gas volume fractions. On the
one hand we can observe in Fig. 9a that the cluster size and growth
rate is the same for the two bubble diameters, suggesting that the
size of the clusters is mainly determined by the uid properties,
channel width and the amount of available gas. On the other hand,
we veried (Fig. 9b) that the number of bubbles embedded in the
cluster is larger for the case of the small bubbles than the larger
ones. Additionally, we found that the normalized standard deviation of the cluster size (/d b ) is larger for the small bubble (up to
123) in comparison to the large bubble case (up to 5), indicating
that the clusters made with the small bubbles interchange bubbles
more often with their surroundings.

SW

J. Rodrigo Vlez-Cordero, R. Zenit / J. Non-Newtonian Fluid Mech. 166 (2011) 3241

39

not necessarily induces bubble clustering. We found that a combination of the values of the Reynolds and Etvs numbers give the
conditions for cluster formation. The value of the ow index is taken
into account in the apparent viscosity included in the Re number.
This kind of dependence of the forces acting on the bubbles with
the ow index is in agreement with the results of Zhu et al. [34],
who work with thinning uids that were forced to pass through a
xed arrangement of rigid spheres. They found that the drag force
experimented on a test particle does not depend on the values of
the ow index but in the Reynolds number and the spacing between
particles.
To identify the hydrodynamic conditions for which bubble clusters are formed we show our results in terms of the Reynolds and
Etvs numbers for the single bubble cases (using Fig. 4), the Eo
(also known as Bond) number being dened as:
Eo =

2
Fig. 10. Bubble velocity variance Tb divided by U SW
as a function of g . Experimental

values of this work: () db = 2.1 mm, n = 0.85; () d b = 3.1 mm, n = 0.85; () d b =
4.2 mm, n = 0.85; () d b = 2.1 mm, Newtonian uid. Other experimental values: ()

data taken from [8] with Newtonian uids, ( ) data taken from [17].

liquid phase was forced to ow in a co-current manner with the

gas phase. Cartellier et al. [43] observed that the dispersion was
composed by free bubbles, bubble pairs, some triplets and clusters composed by more than ve bubbles, and also reported a gas
concentration increase at the center of the column. As the volumetric gas ow rate was further increased, the gas fraction prole
gained the classical saddle shape, that is, with wall peaked distributions, and bubble aggregates were dissociated. Although in the
present work the types of bubble associations mentioned by Cartel SW /USI
lier et al. [43] in a Newtonian uid was not observed, the U
ratios obtained with this uid with the smallest and medium bubble diameters were higher than one (up to 1.3) in the limit of zero
gas fraction, contrary to that observed by Zenit et al.[17], reveling
a kind of long distance interaction between bubbles. In addition
to this, the bubble velocity variance measured for the Newtonian
uid at low Re numbers (0.6 < Re < 1.1) also surpassed by nearly one
order of magnitude the values of more inertia dominated ows [8]
(see Fig. 10).
From all this evidence we can argue that in viscous dominated
bubbly ows (in both Newtonian and thinning uids) the long range
interactions promote the increase of the rise velocity of the bubbles comparing with the single bubble value. This kind of behavior is
similar to the one seen with settling particles [20,21]. Other effects
of these viscous ows could be the increase of bubble uctuations,
precisely due to the long range bubble interactions, and a net repulsion force between the bubbles and the column walls [44], which
could explain the abnormal gas fraction proles obtained by Cartellier et al. [43]. Clustering in thinning uids has, however, important
differences with the one observed in Newtonian uids: the number
of bubbles embedded in the cluster is higher in comparison with
the number of bubbles found in the Newtonian case; the compactness of the clusters is also higher in the case of the thinning uids.
We think that cluster growth in thinning uids is a mechanism
that reduces the spatial viscosity differences in the ow leading to
a Newtonian-like state as the gas fraction is increased. This idea is
explained in the companion paper [23].
5. Conditions for cluster formation
As we saw in Section 4.4, the increase of the velocity ratio
SW /USI , due to bubble clustering, does not depend directly on the
U
value of the ow index n. This means that a more thinning uid does

gdb2


(7)

For the case of isolated bubbles the EoRe plot is often used to
identify the shapes of the bubbles [45]. Cluster formation can be
identied by direct observation of the bubbly ow and also by eval SW /USI with the gas volume fraction. When
uating the change of U
the velocity ratio is larger than 1.5, cluster formation was observed.
In Fig. 8a, for example, bubble clusters are formed in the n = 0.5 and
0.7 uids; in Fig. 8c all the thinning uids formed bubble clusters
SW /USI = 1.5 only serves as a dissince d b was small. The relation U
tinction between the condition where the velocity of the bubble
swarms was higher than the single velocity but without the formation of clusters, as in the Newtonian uid. We identied cluster
formation mainly by simple observation. In the case of the n = 0.32
uid with the smaller bubbles (2.9 mm), cluster identication was
not easy at gas fractions around 0.4%. At this value clusters become
progressively diluted by the whole ow. Nevertheless, we recognized these ows as part of the clustering condition since the values
SW /USI were above 1.5.
of U
Another parameter that can be used to recognize bubble clustering is the bubble velocity variance. As shown in Section 4.6, bubble
clustering increases considerably the velocity variance of the liquid
and bubbles. In our bubbly ows the normalized standard deviation

2 was about 0.4 in unclustered ows while it reached up
Tb /U SW
to 5.0 when clusters were formed.
The result of this mapping is shown in Fig. 11; it includes the
ve uids used in this work. We can describe this plot as follows:
the data is separated according to the clustering behavior, lled
and empty symbols show ow conditions where clustering was or
was not detected, respectively. For instance, a lled symbol was
assigned to a EoRe point if for the same bubble diameter and liquid properties the bubbles clustered for a particular gas volume
fraction. This mapping is easy to do because the gradual increase of
the gas fraction produce only a slight increase of d b . A separation of
the two regimes can be observed. Clearly, the clustering is observed
when the viscous effects are more important than the inertial ones
(small Re). Hence, as inertia increases it is possible to break the clustering behavior. Additionally, the deformability of the bubbles is
also important. For a given value of Re, a ow with more deformable
bubbles will tend to cluster more easily (as the Eo increases). The
separation between the two regimes is given approximately by the
iso-Morton line of 1 103 , the Morton number being dened as:
Mo =

g4
 3

(8)

Previously, a Morton value of 4 104 was proposed [9,46] as

a transition indicator from non-coalescing ows to coalescing
ows in Newtonian uids. This suggest that the transition of noninteracting to interacting bubbles is a general condition for inelastic
uids, including the Newtonian and shear-thinning uids. In this

40

J. Rodrigo Vlez-Cordero, R. Zenit / J. Non-Newtonian Fluid Mech. 166 (2011) 3241

10

in the thinning uids which grow with the gas fraction and increase
the mean bubble velocity with respect to the velocity achieved
by single bubbles. Such clusters, which actually lead to contact
between bubbles, have larger dimensions and lifetime than the
ones observed in Newtonian potential ows [16,17]. By measur SW at different
ing the mean rise velocity of the bubble swarms U
gas fractions, and non-dimensionalized them with the single bubble velocities USI , comparison with previous theoretical data could
be conducted. The main conclusions of this investigation are:

FREE BUBBLES
n=0.3
n=0.8
1

Reynolds

10

10

Mo=10

n=0.7
3

10

Mo=10

10

CLUSTERING
n=0.5

10

10

Eotvos
Fig. 11. Cluster condition formation mapped in a EoRe curve of the single bubbles.
() hydrodynamic conditions for bubble clustering, () hydrodynamic conditions for
free bubbles, ( ) Mo = 103 , (. . .) Mo = 104 , () Newtonian values. The iso-Morton
lines were taken from [45].

work no clustering was seen at rst glance in the Newtonian uid.

Nevertheless, a pair of bubbles rising in a viscous Newtonian uid
can form a stable doublet (see the companion paper [23]). This
indicates that the gas fraction has an important role in cluster
desegregation in Newtonian uids, as seen by Cartellier et al. [43].
The Mo number increases from top to bottom in a EoRe plot.
The tendency to form bubble clusters increases with the Morton
number. That is why the n = 0.55 uid, with the highest Mo number,
SW /USI values (see Fig. 8) followed by the n = 0.76
has the biggest U
uid and nally by the n = 0.85 and 0.32 uids.
We could infer from Fig. 11 that at very low Re numbers, in
the creeping ow regime, the curves displayed in Fig. 8 will have
the same trend as the theoretical ones, that is to say, the n = 0.32
SW /USI values, as it will have the highest
uid will have the highest U
Morton number, followed by the n = 0.55 and so on. As discussed by
other authors [47,46], bubble interaction strongly depends on the
bubbles wake and vortices structure, which in turn evolve as the Re
number is increased or the bubble shape is changed. Additionally,
the degree of deformation of a bubble, measured with the Etvs
(or Bond) number is another important factor that enhance bubble alignment and coalescence, as explained by Manga and Stone
[48,49]. In this work a maximum Eo number of 10 was reached.
It would be interesting to investigate if bubbles with Eo > 40 and
Mo > 1 103 , in which according to [45] the skirted and dimpled
ellipsoidal-cap bubbles are observed, also form clusters in thinning
uids.
In the companion paper of this investigation [23] we present a
detailed study of the interaction of two bubbles rising in the same
shear-thinning uids employed in this work. The results, which
pointed out the importance of the shear-thinning wake formed by a
leading bubble, complement and explain the appearance of bubble
clusters in thinning uids whose structure differs signicantly from
the ones observed in Newtonian bubbly ows.

SW /USI as a func1. Qualitative agreement of the velocities ratio U

tion of the gas volume fraction was found with the theoretical
studies [28,35]. Therefore, the Happel cell model can predict the
effects of the reduction of the local viscosity together with the
hydrodynamic hindrance provoked by bubble interactions.
2. Quantitative agreement with the velocities ratio was not found
between the experiments of this work and the theoretical pre SW /USI was
dictions since the maximum of the velocity ratio U
found at lower gas fractions than the ones predicted by the theory, which this work revealed that is was incomplete. In fact,
the permissible experimental gas fraction values are much lower
than the theoretical ones for the cases of mono-dispersed bubbly ows. The heterogeneous or churn-turbulent regime has
interesting issues by itself but cannot be used to compare experimental results with the available theoretical data.
SW /USI was higher than the
3. The magnitude of the velocity ratio U
theoretical one even at low values of the gas fraction. The difference is attributed to the formation of clusters which the Happel
cell model does not account for.
4. The ratio of the mean bubble velocity and the single bubble
velocity did not follow an orderly correspondence with the ow
index values, as seen in the theoretical works. Instead of this,
SW /USI values could be related to the values of the Reynolds
the U
and Etvs numbers of the single bubbles. Moreover, the bubble
clustering condition was mapped in a EoRe plot. Two regimes
were identied: a free bubble regime and a cluster formation
regime. The limit between one and the other is close to a critical
Morton number (4 104 ) which has been identied as a transition indicator from non-coalescing ows to coalescing ows in
Newtonian uids.
We should mention that although the EoRe plot proved to be
useful in predicting bubble clustering using single bubble data, it
could underestimate the roll that the gas fraction has on cluster
growth. The inuence of the gas fraction, which seems to be important in Newtonian uids and lesser in shear-thinning uids, should
be claried in the future.
Measurements were also conducted to obtain the equivalent
diameter of the bubble cluster and bubble velocity variance. These
results suggest that the cluster growth is not related to the size
of individual bubbles and that the dimensionless bubble velocity
2 increases signicantly when bubble clusters are
variance Tb /U SW
formed.
It is our hope that this work will contribute to the understanding
of the formation of bubble clusters in shear-thinning uids, which
is an important issue in multiphase ows as well as a precursor to
bubble coalescence and change of regime in bubble ows.

6. Conclusions
Acknowledgements
In this work experiments were conducted to study cluster formation in bubbly ows with shear-thinning uids. Special care
was taken to produce nearly mono-dispersed bubbly ows and to
employ thinning uids with negligible elasticity. The visual observation of the ow revealed the formation of dense bubble clusters

The authors will like to thank the support of G. Snchez, F.

Calderas and L. Medina for their help in the rheological measurements. R. Vlez acknowledges CONACyT-Mxico for its nancial
support during his doctoral studies.

J. Rodrigo Vlez-Cordero, R. Zenit / J. Non-Newtonian Fluid Mech. 166 (2011) 3241

References
[1] N. Kantarci, F. Borak, K.O. Ulgen, Bubble column reactors, Proc. Biochem. 40
(2005) 2263.
[2] W.-D. Deckwer, Bubble Column Reactors, Wiley, NY, 1992.
[3] Y.T. Shah, B.G. Kelkar, S.P. Godbole, W.-D. Deckwer, Design parameters estimation for bubble column reactors, A. I. Ch. E. J. 28 (1982) 353.
[4] M. Zlokarnik, Trends and needs in bioprocess engineering, Chem. Eng. Prog. 86
(1990) 62.
[5] H.W. Blanch, S.M. Bhavaraju, Rheology of fermentation broths, Biotechnol. Bioeng. 18 (1976) 745.
[6] P.M. Kilonzo, A. Margaritis, The effects of non-Newtonian fermentation broth
viscosity and small bubble segregation on oxygen mass transfer in gas-lift
bioreactors: a critical review, Biochem. Eng. J. 17 (2004) 27.
[7] L. van Wijngaarden, On pseudo turbulence, Theor. Comput. Fluid Dynam. 10
(1998) 449.
[8] J. Martnez-Mercado, C.A. Palacios-Morales, R. Zenit, Measurement of pseudoturbulence intensity in monodispersed bubbly liquids for 10 < Re < 500, Phys.
Fluids 19 (2007) 103302.
[9] A. Smolianski, H. Haario, P. Luukka, Numerical study of dynamics of a single
bubbles and bubbles swarm, Appl. Math. Model. 23 (2008) 641.
[10] N. Kishore, R.P. Chhabra, V. Eswaran, Drag on ensembles of uid spheres translating in a power-law liquid at moderate Reynolds numbers, Chem. Eng. J. 139
(2008) 224.
[11] S. Radl, G. Tryggvason, J.G. Khinast, Flow and mass transfer of fully resolved
bubble in non-Newtonian uids, A. I. Ch. E. J. 53 (2007) 1861.
[12] R.G. Larson, The Structure and Rheology of Complex Fluids, Oxford University
Press, 1999.
[13] H.H. Hu, N.A. Patankar, M.Y. Zhu, Direct numerical simulations of uidsolid
systems using the arbitrary LagrangianEulerian technique, J. Comput. Phys.
169 (2001) 427.
[14] G. Gheissary, B.H.A.A. van den Brule, Unexpected phenomena observed in particle settling in non-Newtonian media, J. Non-Newton. Fluid Mech. 67 (1996)
1.
[15] D.D. Joseph, Y.J. Liu, M. Poletto, J. Feng, Aggregation and dispersion of spheres
falling in viscoelastic liquids, J. Non-Newton. Fluid Mech. 54 (1994) 45.
[16] B. Figueroa-Espinoza, R. Zenit, Clustering in high Re monodispersed bubbly
ows, Phys. Fluids 17 (2005) 091701.
[17] R. Zenit, D.L. Koch, A.S. Sangani, Measurements of the average properties of a
suspension of bubbles rising in a vertical channel, J. Fluid Mech. 429 (2001)
307.
[18] H. Buchholz, R. Buchholz, J. Lcke, K. Schgerl, Bubble swarm behavior and
gas absortion in non-Newtonian uids in sparged columns, Chem. Eng. Sci. 33
(1978) 1061.
[19] A. Schumpe, W.-D. Deckwer, Gas holdups, specic interfacial areas, and mass
transfer coefcients of aerated carboxymethyl cellulose solutions in a bubble
column, Ind. Eng. Chem. Proc. Des. Dev. 21 (1982) 706.
[20] K.O.L.F. Jayaweera, B.J. Mason, G.W. Slack, The behavior of cluster of spheres
falling in a viscous uid, J. Fluid Mech. 20 (1964) 121.
[21] L.M. Hocking, The behavior of clusters of spheres falling in a viscous uid, J.
Fluid Mech. 20 (1964) 129.
[22] D.D. Joseph, Y.J. Liu, Orientation of long bodies falling in a viscoelastic liquid, J.
Rheol. 37 (1993) 961.
[23] J.R. Vlez-Cordero, D. Samano, P. Yue, J.J. Feng, R. Zenit, Hydrodynamic interaction between a pair of bubbles ascending in shear-thinning inelastic uids, J.
Non-Newton. Fluid Mech., submitted for publication.
[24] W.-D. Deckwer, K. Nguyen-tien, A. Schumpe, Y. Serpemen, Oxygen mass transfer into aerated CMC solutions in a bubble column, Biotechol. Bioeng. 24 (1982)
461.

41

[25] S.P. Godbole, A. Schumpe, Y.T. Shah, N.L. Carr, Hydrodynamics and mass transfer in non-Newtonian solutions in a bubble column, A. I. Ch. E. J. 30 (1984)
213.
[26] N.W. Haque, K.D.P. Nigam, K. Viswanathan, J.B. Joshi, Studies on gas holdup
and bubble parameters in bubble columns with (carboxymethyl)cellulose solutions, Ind. Eng. Chem. Res. 26 (1987) 86.
[27] N.W. Haque, K.D.P. Nigam, K. Viswanathan, J.B. Joshi, Studies on bubble rise
velocity in bubble columns employing non-Newtonian solutions, Chem. Eng.
Commun. 73 (1988) 31.
[28] S. Gummalam, R.P. Chhabra, Rising velocity of a swarm of spherical bubbles in
a power law non-Newtonian liquid, Can. J. Chem. Eng. 65 (1987) 1004.
[29] S.M. Bhavaraju, R.A. Mashelkar, H.W. Blanch, Bubble motion and mass transfer
in non-newtonian uids: Part II. Swarm of bubbles in a power law uid, A. I.
Ch. E. J. 24 (1978) 1070.
[30] T. Hirose, M. Moo-Young, Bubble Drag and Mass Transfer in non-Newtonian
Fluids: Creeping Flow with Power-law Fluids, Can. J. Chem. Eng. 47 (1969) 265.
[31] J. Happel, Viscous ow in multiparticle systems: slow motion of uids relative
to beds of spherical particles, A. I. Ch. E. J. 4 (1958) 197.
[32] D. Rodrigue, D. De Kee, C.F. Chan Man Fong, A note on the drag coefcient of a
single gas bubble in a power-law uid, Can. J. Chem. Eng. 77 (1999) 766.
[33] A. Acharya, R.A. Mashelkar, J. Ulbrecht, Mechanics of bubble motion and deformation in non-newtonian media, Chem. Eng. Sci. 32 (1977) 863.
[34] C. Zhu, K. Lam, H.-H. Chu, X.-D. Tang, G. Liu, Drag forces of interacting spheres
in power-law uids, Mech. Res. Commun. 30 (2003) 651.
[35] R.P. Chhabra, Rising velocity of a swarm of spherical bubbles in power law uids
at high Reynolds numbers, Can. J. Chem. Eng. 76 (1998) 137.
[36] P.D.M. Spelt, A.S. Sangani, Properties and averaged equations for ows of bubbly
liquids, Appl. Sci. Res. 58 (1998) 337.
[37] H.N. Oguz, A. Prosperetti, Dynamics of bubble growth and detachment from a
needle, J. Fluid Mech. 257 (1993) 111.
[38] R.R. Lessard, S.A. Zieminski, Bubble coalescence and gas transfer in aqueous
electrolytic solutions, Ind. Eng. Chem. Fundam. 10 (1971) 260.
[39] F. Calderas, A. Sanchez-Solis, A. Maciel, O. Manero, The transient ow of the PETPEN-montmorillonite clay nanocomposite, Macromol. Symp. 283284 (2009)
354.
[40] E. Soto, C. Goujon, R. Zenit, O. Manero, A study of velocity discontinuity for single
air bubbles rising in an associative polymer, Phys. Fluids 18 (2006) 121510.
[41] M. Ishii, N. Zuber, Drag coefcient and relative velocity in bubbly, droplet or
particulate ows, A. I. Ch. E. J. 25 (1979) 843.
[42] A. Cartellier, N. Rivire, Bubble-induced agitation and microstructure in uniform bubbly ows at small to moderate particle Reynolds numbers, Phys. Fluids
13 (2001) 2165.
[43] A. Cartellier, L. Timkin, N. Rivire, New structures of Poiseuille bubbly ows due
to clustering, in: Proc. ASME-FEDSM 97, paper 3528, Vancouver, June 2226,
1997.
[44] P. Vasseur, R.G. Cox, The lateral migration of spherical particles sedimenting in
a stagnant bounded uid, J. Fluid Mech. 80 (1977) 561.
[45] R. Clift, J.R. Grace, M.E. Weber, Bubbles, Drops, and Particles, Academic Press,
NY, 1978.
[46] C.W. Stewart, Bubble interaction in low-viscosity liquids, Int. J. Multiphase Flow
21 (1995) 1037.
[47] D. Legendre, J. Magnaudet, G. Mougin, Hydrodynamic interactions between
two spherical bubbles rising side by side in a viscous liquid, J. Fluid Mech. 497
(2003) 133.
[48] M. Manga, H.A. Stone, Buoyancy-driven interactions between two deformable
viscous drops, J. Fluid Mech. 256 (1993) 647.
[49] M. Manga, H.A. Stone, Collective hydrodynamics of deformable drops and bubbles in dilute low Reynolds number suspensions, J. Fluid Mech. 300 (1995)
231.